G-quadruplexes are transcription factor binding hubs in human chromatin

Abstract

Background: The binding of transcription factors (TF) to genomic targets is critical in the regulation of gene expression. Short, double-stranded DNA sequence motifs are routinely implicated in TF recruitment, but many questions remain on how binding site specificity is governed.

Results: Herein, we reveal a previously unappreciated role for DNA secondary structures as key features for TF recruitment. In a systematic, genome-wide study, we discover that endogenous G-quadruplex secondary structures (G4s) are prevalent TF binding sites in human chromatin. Certain TFs bind G4s with affinities comparable to double-stranded DNA targets. We demonstrate that, in a chromatin context, this binding interaction is competed out with a small molecule. Notably, endogenous G4s are prominent binding sites for a large number of TFs, particularly at promoters of highly expressed genes.

Conclusions: Our results reveal a novel non-canonical mechanism for TF binding whereby G4s operate as common binding hubs for many different TFs to promote increased transcription.

Keywords: Chemical biology; DNA G-quadruplex; Gene expression; Transcription factor binding.

Fig. 1

TF binding is tightly linked to endogenous G4 structures in the human genome. a Enrichment of 524 ENCODE proteins at G4 ChIP-seq sites in K562 for randomization in open chromatin (DHS). Top 20 and bottom 20 candidates are highlighted. Green shading indicates proteins with reported G4-association. b Genomic association with endogenous G4s is consistent in K562 and HepG2 cells. Spearman correlation test (r_s, ****P < 0.0001) is based on the maximum enrichment observed for TFs that have been mapped in both K562 (x-axis) and HepG2 (y-axis) cells. Green shading indicates proteins with reported G4-association. c Genomic association of TFs obtained from ENCODE with endogenous G4s (x-axis) and potential G4 control sites at promoters (y-axis). Proteins for which binding is independent of secondary structure formation should show similar enrichment for both data sets (white dashed line). Green shading indicates proteins with reported G4 association. d Occupancy profiles of enriched candidates SP2, E2F4, FUS, and NRF1 and a non-enriched TF, CBX8, around endogenous G4 sites (green) and control sequences (gray). The strandedness of endogenous G4s was derived from stranded data of sequences with G4 forming potential [26] (see “Materials and methods”)

Fig. 2

TFs selectively bind to G4 structures. a Affinity pull-down western blot analysis of different G4 oligonucleotides and control sequences. Genomic enrichment at endogenous G4s in K562 for randomization in open chromatin is shown in brackets. b Affinity pull-down of SP2, FUS, and NRF1 using a G4 oligomer, single-stranded oligomers unable to form a G4 structure (ss mutMyc and ss Myc*) and respective consensus sequences. c Binding curves as determined by ELISA show high-affinity binding of recombinant FLAG-NRF1 to a NRF1 double-stranded DNA consensus sequence and G4 structures, but significantly weaker binding to a single-stranded 7-deaza control (error bars display standard deviation, N = 3)

Fig. 3

Competition of TF binding to G4s in native chromatin by small molecule ligands. a Competition ELISA. Immobilized G4 Myc and a double-stranded DNA consensus oligomer were pre-incubated with increasing concentrations of G4 ligand PDS followed by recombinant FLAG-NRF1 (20 nM) (error bars display standard deviation, N = 3). b PDS dose-dependent competition for NRF1 in K562 cell nuclear lysates. PDS displaces TFs from different G4 oligomers, but does not interfere with binding to the double-stranded DNA consensus oligomer (error bars display standard deviation, N = 2). c Scheme for TF displacement upon G4 ligand treatment and detection via native ChIP. d Native ChIP-qPCR for G4-associated SP2, NRF1, and FUS binding shows a PDS-dependent signal reduction. x-axis, selected positive regions for G4 ChIP-seq and TF ENCODE ChIP signal and two negative control regions (ESR1, TMCC1) with no G4 and TF ChIP-seq signal (error bars display standard error of the mean, N = 4). e Native ChIP-qPCR of control CTCF and FOXA1 are not displaced by PDS (error bars display standard error of the mean, N = 4). f PDS-dependent signal reduction in native SP2 ChIP-qPCR at two positive regions (error bars display standard error of the mean, N = 3)

Fig. 4

G4s are hubs for the recruitment TFs to enhance transcription. Throughout panels a–c gene endogenous G4 in promoters accessible in open chromatin (− 1 kb upstream TSS, DHS positive) are colored in green whereas promoters lacking an endogenous G4 are represented in gray. a Endogenous G4s mark genomic regions that are highly occupied by TFs. Proportion of G4s overlapping with multiple different TFs in K562 cells (top) and HepG2 cells (bottom). b Distributions of transcript levels split by the number of TFs binding at G4s in promoters or at promoters lacking G4s in K562 (top) and HepG2 cells (bottom) (unpaired Wilcoxon test). The number of cases (shown in brackets) for higher TF occupancy is substantially higher for G4s. c The average transcriptional output (displayed in transcripts per million (TPM), log₁₀ scale) is compared for genes with and without endogenous G4s in promoters in K562 (left) and HepG2 cells (right) (unpaired Wilcoxon test). d A model for how endogenous G4s can enhance occupancy by multiple TFs at promoters: (i) Repressed promoters are unoccupied by TFs. (ii) Double-stranded DNA consensus binding sites recruit particular TFs to promoters resulting in active transcription. (iii) G4s can recruit numerous different TFs causing even more actively transcribed genes